[Openmp-commits] [openmp] r361421 - [OPENMP][NVPTX]Fix barriers and parallel level counters, NFC.

[Alexey Bataev via Openmp-commits]

Author: abataev
Date: Wed May 22 12:50:32 2019
New Revision: 361421

URL: http://llvm.org/viewvc/llvm-project?rev=361421&view=rev
Log:
[OPENMP][NVPTX]Fix barriers and parallel level counters, NFC.

Summary:
Parallel level counter should be volatile to prevent some dangerous
optimiations by the ptxas. Otherwise, ptxas optimizations lead to
undefined behaviour in some cases.
Also, use __threadfence() for #pragma omp flush and if the barrier
should not be used (we have only one thread in the team), still perform
flush operation since the standard requires implicit flush when
executing barriers.

Reviewers: gtbercea, kkwli0, grokos

Subscribers: guansong, jfb, jdoerfert, openmp-commits, caomhin

Tags: #openmp

Differential Revision: https://reviews.llvm.org/D62199

Modified:
    openmp/trunk/libomptarget/deviceRTLs/nvptx/src/omp_data.cu
    openmp/trunk/libomptarget/deviceRTLs/nvptx/src/omptarget-nvptx.h
    openmp/trunk/libomptarget/deviceRTLs/nvptx/src/sync.cu

Modified: openmp/trunk/libomptarget/deviceRTLs/nvptx/src/omp_data.cu
URL: http://llvm.org/viewvc/llvm-project/openmp/trunk/libomptarget/deviceRTLs/nvptx/src/omp_data.cu?rev=361421&r1=361420&r2=361421&view=diff
==============================================================================
--- openmp/trunk/libomptarget/deviceRTLs/nvptx/src/omp_data.cu (original)
+++ openmp/trunk/libomptarget/deviceRTLs/nvptx/src/omp_data.cu Wed May 22 12:50:32 2019
@@ -31,7 +31,8 @@ __device__ omptarget_nvptx_SimpleMemoryM
 __device__ __shared__ uint32_t usedMemIdx;
 __device__ __shared__ uint32_t usedSlotIdx;
 
-__device__ __shared__ uint8_t parallelLevel[MAX_THREADS_PER_TEAM / WARPSIZE];
+__device__ __shared__ volatile uint8_t
+    parallelLevel[MAX_THREADS_PER_TEAM / WARPSIZE];
 __device__ __shared__ uint16_t threadLimit;
 __device__ __shared__ uint16_t threadsInTeam;
 __device__ __shared__ uint16_t nThreads;

Modified: openmp/trunk/libomptarget/deviceRTLs/nvptx/src/omptarget-nvptx.h
URL: http://llvm.org/viewvc/llvm-project/openmp/trunk/libomptarget/deviceRTLs/nvptx/src/omptarget-nvptx.h?rev=361421&r1=361420&r2=361421&view=diff
==============================================================================
--- openmp/trunk/libomptarget/deviceRTLs/nvptx/src/omptarget-nvptx.h (original)
+++ openmp/trunk/libomptarget/deviceRTLs/nvptx/src/omptarget-nvptx.h Wed May 22 12:50:32 2019
@@ -398,7 +398,7 @@ extern __device__ omptarget_nvptx_Simple
     omptarget_nvptx_simpleMemoryManager;
 extern __device__ __shared__ uint32_t usedMemIdx;
 extern __device__ __shared__ uint32_t usedSlotIdx;
-extern __device__ __shared__ uint8_t
+extern __device__ __shared__ volatile uint8_t
     parallelLevel[MAX_THREADS_PER_TEAM / WARPSIZE];
 extern __device__ __shared__ uint16_t threadLimit;
 extern __device__ __shared__ uint16_t threadsInTeam;

Modified: openmp/trunk/libomptarget/deviceRTLs/nvptx/src/sync.cu
URL: http://llvm.org/viewvc/llvm-project/openmp/trunk/libomptarget/deviceRTLs/nvptx/src/sync.cu?rev=361421&r1=361420&r2=361421&view=diff
==============================================================================
--- openmp/trunk/libomptarget/deviceRTLs/nvptx/src/sync.cu (original)
+++ openmp/trunk/libomptarget/deviceRTLs/nvptx/src/sync.cu Wed May 22 12:50:32 2019
@@ -62,6 +62,8 @@ EXTERN void __kmpc_barrier(kmp_Ident *lo
         // Barrier #1 is for synchronization among active threads.
         named_sync(L1_BARRIER, threads);
       }
+    } else {
+      __kmpc_flush(loc_ref);
     } // numberOfActiveOMPThreads > 1
     PRINT0(LD_SYNC, "completed kmpc_barrier\n");
   }
@@ -130,7 +132,7 @@ EXTERN void __kmpc_end_single(kmp_Ident
 
 EXTERN void __kmpc_flush(kmp_Ident *loc) {
   PRINT0(LD_IO, "call kmpc_flush\n");
-  __threadfence_system();
+  __threadfence();
 }
 
 ////////////////////////////////////////////////////////////////////////////////